Multilevel Memory Programming with Variable Conductance Windows
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multilevel memory devices face a tradeoff between increasing memory states for higher data storage density and maintaining good memory retention, as larger memory windows enhance retention but reduce the number of states, while smaller windows increase the risk of retention failure.
Innovation Solution
Implementing variable memory windows by assigning larger windows to low-conductance states with high variability and smaller windows to high-conductance states with low variability, optimizing the separation between adjacent conductance states for improved retention and memory states.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If larger memory windows are used to enhance retention, then reliability is improved, but the number of memory states decreases
Solution Approach 1:
The patent applies local quality by assigning different memory window sizes to different conductance states based on their specific needs. Low-conductance states with high variability are assigned larger memory windows to ensure adequate separation and retention, while high-conductance states with low variability are assigned smaller memory windows to maximize the number of states. This localized optimization resolves the contradiction by making the memory window size adaptive to the local characteristics of each conductance state rather than using a uniform size.
Solution Approach 2:
The patent changes the parameter of memory window size from a fixed uniform value to a variable value that depends on the conductance state. By calculating and applying different window sizes based on the variability characteristics of each state, the system optimizes both retention (for states needing larger windows) and state density (for states that can use smaller windows), thereby resolving the contradiction between reliability and quantity of memory states.
2Quantity of substance
If smaller memory windows are used to increase the number of memory states, then data storage density is improved, but the risk of retention failure increases
Solution Approach 1:
The patent applies local quality by making the memory window size adaptive to the local variability characteristics of each conductance state. States with low variability can use smaller windows, while states with high variability receive larger windows. This localized adaptation allows the system to increase the overall number of states while maintaining adequate retention for each individual state, thus resolving the contradiction between density and reliability.
Solution Approach 2:
The patent changes the memory window parameter from a uniform fixed value to a variable value that is calculated based on the variability of each conductance state. This parameter transformation enables the system to optimize the balance between having enough states for high density and maintaining sufficient separation for reliable retention, directly resolving the contradiction.
3Device complexity
If uniform memory windows are used, then device complexity is reduced, but the optimization of separation between conductance states is compromised
Solution Approach 1:
The patent applies local quality by assigning memory window sizes tailored to the specific characteristics of each conductance state rather than using a uniform size. This approach accepts increased complexity in window assignment as necessary to achieve optimal separation between states, particularly for states with high variability that require larger windows to maintain adequate separation.
Solution Approach 2:
The patent introduces dynamics by making the memory window size adaptive and variable based on the conductance state characteristics. Rather than a static uniform window size, the system dynamically adjusts window sizes to match the needs of each state, optimizing separation while accepting the increased complexity as a necessary trade-off for performance.
Data Source
AI summary
The present disclosure provides methods for programming multilevel memory devices. The methods may include determining a first plurality of memory windows representative of gaps between dispersions of adjacent conductance states of the memory device, determining a plurality of dispersion parameters representative of estimated dispersions of the conductance states, and determining a second plurality of memory windows based on the first plurality of memory windows and the plurality of dispersion parameters. The second plurality of memory windows represents separations between mean conductance values of adjacent conductance states of the memory device. The second plurality of memory windows has varying values.